| In recent years, general, commercial and military aviation has been plagued by the occurrence of several crashes, in which faulty control surfaces appeared to be the cause, or one of the possible causes. Typically, the National Transportation and Safety Board (NTSB) begins its investigation by analyzing the wreckage along with the flight data recorder (FDR) and cockpit voice recorder (CVR). In most instances this combination of information is enough for the NTSB to discover the cause of the crash, but unfortunately this is not always the case. From this the argument can be made that there is a need for a scheme which can accurately predict and/or estimate additional aircraft time histories to aid investigators in the unfortunate event of a crash. The most critical parameters of interest to investigators, not recorded by the FDRs, are the primary control surface deflections (aileron, rudder, and elevator). These additional parameters give investigators great insight into the aircraft dynamics in those critical moments before the crash. Therefore the purpose of this research effort is to develop and verify, through a series of tests, a scheme which can accurately reconstruct the aircraft's primary surface deflection using data available from the FDR. The reconstruction scheme consists of two neural networks. The first is used to simulate the aircraft dynamics, while the second is used to reconstruct the primary surface deflection, and or system input. The scheme was then verified by applying it to several non-linear models including a general SISO system, a general MIMO system, and a non-linear model of an F-16 Fighting Falcon. In each case the reconstruction scheme proved successful in reconstructing the unknown inputs or surface deflections given only the dynamics of the system. |
